The combination of high resolution assays in genomics with microscopic imaging has been used for the detection of complex chromosomal rearrangements, a significant but difficult problem in prenatal and postnatal diagnosis, birth defect detection and cancer research. As a recently developed molecular cytogenetic technique, multiplex fluorescence in situ hybridization (M-FISH) imaging has provided rapid and high resolution detection of chromosomal abnormalities associated with cancer and genetic disorders. However, the technique is currently limited to research use and only serves as an adjunct tool to the G-banding based monochromatic chromosomal karyotyping in a clinical laboratory. A primary barrier of the technique is the lower classification accuracy when classifying chromosomes from multi-color microscopic imaging data. Therefore, the goal of this R15 project is to develop innovative multi-spectral image processing and machine learning techniques for M-FISH image analysis so that chromosomal rearrangement detection can be made more reproducible, robust, and faster, thereby significantly increasing the ability and efficacy of this newly developed cellular imaging technique. Our proposed approaches such as multiscale feature extraction, nonlinear manifold analysis and adaptive fuzzy clustering are able to target specific features of multi-spectral imaging data, promising a significant improvement over the current techniques. In order to validate the technique and bring it into clinical use, we will partner with a clinical geneticist, Dr. Merlin Butler, and a cytogeneticist, Dr. Diane Persons both at Kansas University Medical Center. In addition, we will collaborate with an industrial scientist, Dr. Kenneth Castleman, who is the pioneer in developing and commercializing cytogenetic imaging products. Through our interdisciplinary research and collaboration, we will accomplish the following specific aims: 1) develop image normalization approaches to improve the acquisition of multi-color FISH images;2) develop multiscale dimension analysis to extract features from multi-color images;3) design adaptive fuzzy clustering and incorporate contextual information to improve the pixel-wise classification of chromosomes;and 4) validate computational approaches with clinical testing in collaboration with medical and industrial partners. This research project will also enhance our research infrastructure in biomedical image informatics and provide undergraduate and graduate students opportunities to touch the frontier of molecular and cellular imaging by participating in the proposed research activities. PUBLIC HEALTH RELEVANCE: Unraveling complex rearrangements using cytogenetic approaches such as M-FISH imaging has been extremely useful in prenatal, postnatal and cancer diagnoses. Our proposed approaches have the potential to significantly improve the reliability of the newly developed M-FISH imaging technique, making it feasible for clinical use. This will in turn benefit the health of human beings. Furthermore, the developed computational techniques can be applicable to a wide range of multi-color bio-imaging problems, thereby having a broad impact on the biomedical community.